Erlang ANN

Or: ANNie Get Your Neural Net

Artificial Neural Networks (ANNs) are a class of mathematical models based on ideas taken from biological systems like the brain. They model classification problems by taking a large number of fairly dumb units (perceptrons) and adjusting the connections between them to identify patterns. The effect of this is to map input elements into n-dimensions and then combine these in such a way that an acceptable output is found. The key to these systems is that the connections can adapt and converge on a proper mapping through a training and learning phase.

The goal of Erlang ANN is to serve as an educational project on simple neural nets and Erlang OTP. I am modifying the basic framework created in http://www.trapexit.org/Erlang_and_Neural_Networks to make use of the OpenTelecom Platform. I hope to also create a system that can be applied to some classification tasks.

Design

This is an educational experiment for myself. I will update and use this section to describe the architecture of the system. I have no plans for writing it out as a tutorial, but I will try to pack my commit messages with meaty details about the chronological decisions and this README with the current or intended overall design.